Skip to content

Benchmark

Dtbench EDT answer preference

Coding benchmarks compare models on programming tasks. Scores are percentages.

Models measured
170
Scores recorded
177
Sources
1
IQ weight
0.99

Score distribution

67.8%
Mean
69%
Median
11.3%
Std dev
53%
10th percentile
81.6%
90th percentile
43.3%
Min
43.3%96.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.99 for chat models, measured over 170 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
101GPT 4· Epoch AI Benchmarking Hub data66.7%
102Mistral Large 3· Epoch AI Benchmarking Hub data65.3%
103Qwen3 30B A3B2507, thinking· Epoch AI Benchmarking Hub data65.3%
104Mistral Medium 3· Epoch AI Benchmarking Hub data64.9%
105Qwen3 30B A3B· Epoch AI Benchmarking Hub data64.5%
106o1high· Epoch AI Benchmarking Hub data64.4%
107Gemma 3 12B IT· Epoch AI Benchmarking Hub data64%
108GPT-6 Lunamax· Epoch AI Benchmarking Hub data64%
109Mistral Small2506· Epoch AI Benchmarking Hub data63.7%
110Claude Haiku 3.5· Epoch AI Benchmarking Hub data63.4%
111GPT 5.4xhigh· Epoch AI Benchmarking Hub data63.2%
112Mistral Small2503· Epoch AI Benchmarking Hub data62.8%
113DeepSeek V4 Flash 0731max· Epoch AI Benchmarking Hub data62.7%
114Mistral Medium2604· Epoch AI Benchmarking Hub data62.7%
115Mistral Small2603· Epoch AI Benchmarking Hub data62.7%
116o3 Minihigh· Epoch AI Benchmarking Hub data62.7%
117Mistral Medium 3.1· Epoch AI Benchmarking Hub data62.2%
118Qwen 3.8 27B· Epoch AI Benchmarking Hub data61.8%
119Qwen2.5 72B Instruct· Epoch AI Benchmarking Hub data61.3%
120Magistral Small 2509· Epoch AI Benchmarking Hub data61.1%
121GPT OSS 120Bhigh· Epoch AI Benchmarking Hub data60.5%
122Mistral Large2407· Epoch AI Benchmarking Hub data60.1%
123Llama 4 Maverick· Epoch AI Benchmarking Hub data60%
124Muse Spark 1.1high· Epoch AI Benchmarking Hub data60%
125GPT-4 Turbo· Epoch AI Benchmarking Hub data59.9%
126Claude 3 Opus· Epoch AI Benchmarking Hub data59.6%
127Claude Instant 1.2· Epoch AI Benchmarking Hub data59.5%
128MiniMax M3· Epoch AI Benchmarking Hub data59.2%
129DeepSeek V3.2· Epoch AI Benchmarking Hub data59.2%
130GPT-3.5 Turbo· Epoch AI Benchmarking Hub data58.9%
131Llama 3.1 405B Instruct· Epoch AI Benchmarking Hub data58.5%
132Llama 3 70B Instruct· Epoch AI Benchmarking Hub data58.3%
133GPT-4o mini· Epoch AI Benchmarking Hub data58.2%
134Mistral Large2411· Epoch AI Benchmarking Hub data58.2%
135Mistral Small· Epoch AI Benchmarking Hub data58%
136Inkling FP4xhigh· Epoch AI Benchmarking Hub data57.9%
137Qwen3 14B· Epoch AI Benchmarking Hub data57.3%
138Ministral 3B· Epoch AI Benchmarking Hub data57.1%
139C4AI Command R 08 2024· Epoch AI Benchmarking Hub data56.8%
140Claude 3 Sonnet· Epoch AI Benchmarking Hub data56.6%
141Muse Spark 1.2xhigh· Epoch AI Benchmarking Hub data56.6%
142gemma2-27b· Epoch AI Benchmarking Hub data56.3%
143Gemini 2.5 Flash-Litethinking· Epoch AI Benchmarking Hub data56.2%
144Grok 2· Epoch AI Benchmarking Hub data56.2%
145Claude 2.1· Epoch AI Benchmarking Hub data55.9%
146GPT-4.1 nano· Epoch AI Benchmarking Hub data55.7%
147Gemma 3 27B IT· Epoch AI Benchmarking Hub data55.6%
148gemini-1.5-pro-002· Epoch AI Benchmarking Hub data55.2%
149Gemma 3 4B IT· Epoch AI Benchmarking Hub data54.9%
150Claude 3 Haiku· Epoch AI Benchmarking Hub data54.9%

Showing 101–150 of 177 results.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence)